g

,

,

p

g

p

ll introduce several unsupervised machine learning

orithms such as density estimation and cluster analysis and

monstrate how these algorithms can be used for responsive

ne discovery. A typical responsive gene discovery problem,

, the essential gene discovery problem, will be introduced

d discussed in this chapter. How essential genes can be

covered using unsupervised machine learning algorithms

ll be demonstrated.

logical question — essential gene discovery

genome, genes may play different roles for a cell or an organism

on, for instance for a bacteria to survive or grow under a specific

. A gene that has such a function for a cell or an organism to

or survive is called an essential gene [Gluecksohn-Waelsch,

the function of such a gene is disrupted or disabled, the organism

l may not survive or replicate. Such a gene is also called a lethal

relationship with the bacterial growth. Discovering and

ising these genes is important because their products can be the

r developing drugs to fight diseases [Rancati, et al., 2018; Bartha,

18].

the transposon technology to discover essential genes has been

for many years. Theoretically, in a survived or a well-grown

species, non-essential genes can attract transposon insertions, but

ial gene will attract no transposon insertion. Therefore analysing

me-wise transposon insertion pattern can rely on the selection

mutant [Rubin, et al., 2015; Yang, et al., 2017]. Recently,

g the genome-wise transposon insertion pattern to discover the

genes for a species has been facilitated by the high-throughput

on sequencing technology [Langridge, et al., 2009; Zomer, et al.,

ng, et al., 2017]. The technology can map millions of transposon

s, hence mutants, to a genome in one experiment. It thus makes it